Structure-property relationship of Co2MnSi thin films in response to He+-irradiation.

We investigated the structure-property relationship of Co2MnSi Heusler thin films upon the irradiation with He+ ions. The variation of the crystal structure with increasing ion fluence has been probed using nuclear magnetic resonance (NMR) and transmission electron microscopy (TEM), and associated with the corresponding changes of the magnetic behavior. A decrease of both the structural order and the moment in saturation is observed. Specifically, we detect a direct transition from a highly L21-ordered to a fully A2-disordered structure type and quantify the evolution of the A2 structural contribution as a function of ion fluence. Complementary TEM analysis reveals a spatially-resolved distribution of the L21 and A2 phases showing that the A2 disorder starts at the upper part of the films. The structural degradation in turn leads to a decreasing magnetic moment in saturation in response to the increasing fluence.

Introduction

Highly L21-ordered Heusler compounds (X2YZ) are promising candidates for half-metallic ferromagnetism[1-6] whose 100% spin polarization gives rise to interesting applications in spintronic devices, such as magnetic tunnel junctions (MTJ) or giant magnetoresistance (GMR) devices[7-11]. In combination with its high Curie temperature (985 K), saturation moment (5 μ) and large band gap for the minority spin states (0.4–0.8 eV)[2,8,9,12,13], Co2MnSi is especially well-suited for such applications. Chemical order has strong impact on the properties in general. This relationship is in particular important for half-metallic ferromagnetic materials, since their spin polarisation is downgraded rapidly for certain types of disorder[14-17]. As a consequence, the performance of any application based on the spin polarisation, e.g. GMR devices, will in turn strongly depend on the degree of chemical order of the material[18]. Hence, a control of the structure-property relationship is a crucial factor for their use in high-end applications. The classical route to address and control the structural order and, thus, the properties of such highly functional materials is post-growth annealing, where the annealing temperature and/or annealing time have to be optimized.

Another route to control the structure and magnetic properties of functional materials is the irradiation with light ions, which is known to strongly modify their magnetic properties[19-22]. Studies on Co2MnSi thin films reported improved structural properties upon the irradiation with He+ ions within a certain range of fluences (1 × 1014–5 × 1014 ions/cm2, 30 keV), going along with an improvement of the electronic and magnetic properties[23]. However, the changes in the observed saturation moment were on the order of a few percent while the structural results were extracted from small changes in the ratio of two superlattice reflections (200 and 400), whose absolute intensities were small, such that changes may be easily over/-underestimated.

In this work, we aim to track irradiation-induced changes of the crystallographic order at the local level. To ensure full characterization of the structural order, we applied both local (TEM) as well as volume-integral probing (NMR) of He+-irradiated Co2MnSi thin films. The combination of magnetometry, nuclear magnetic resonance (NMR) and transmission electron microscopy (TEM) allows to get insights into the structure-property relationship, in particular linking both local and macroscopic properties. NMR probes the local environment of the investigated nuclei over the entire sample volume. It has already been shown that NMR is a potential method to resolve the evolution of structural order in functional materials[18,24-35]. By means of 59Co NMR, we observe a degradation of the crystallographic order on a local scale from L21 to A2 order upon increasing He+ ion fluence. We were able to deduce the amount of the respective L21- and A2-ordered phases for each applied fluence. We complement the NMR results by TEM which provides structural information of selected regions of the sample. The evolution of the structure is linked to the macroscopic magnetic properties showing a decrease of the saturated moment in response to the ion irradiation.

Methods

Co2MnSi thin films of 40 nm thickness covered with a 5 nm thick Ta layer were deposited epitaxially on a (100) MgO substrate by magnetron sputtering and annealed at 500 °C. Samples of similar sizes (~7 × 7 mm2) have been cut and have been exposed to a 15 keV He+ ion beam with fluences between 1 × 1013 and 5 × 1015 ions/cm2. The ion beam was incident perpendicular to the film surface. This irradiation has been carried out at ambient temperature at the Ion Beam Center at the Helmholtz-Zentrum Dresden-Rossendorf. The displacements per atom (dpa) have been simulated using the binary collision approximation (Stopping and Range of Ions in Matter (SRIM) package)[36]. According to these simulations, 15 keV He+ ions provide a nearly flat distribution of atomic displacements within the 40 nm thick Co2MnSi film, while causing negligible intermixing with the capping layer and MgO substrate. The simulated dpa within the Co2MnSi film varied from 0.001 to 0.56 for fluences of 1 × 1013 to 5 × 1015 ions/cm2, respectively (see Supplementary Fig. 2 of supplementary material), leading to a statistical variation of the local ordering. NMR is an ideal tool to investigate such statistical variations.

The magnetization has been measured at 300 K using a superconducting quantum interference device (SQUID) from Quantum Design in the field range between −20 to 20 kOe. Please note, that the magnetization measured at 300 K does not significantly differ from the respective 5 K/0 K values due to the high Curie temperature of 985 K[37]. Zero-field (ZF) NMR measurements were performed at 5 K with an automated, coherent, phase-sensitive and frequency-tuned spin-echo spectrometer (NMR Service, Erfurt, Germany). A solid-echo sequence, consisting of two 90° pulses of 0.6 μs width, separated by 5 μs was used to observe the spin echo. This pulse sequence was repeated up to 2000 times to get a satisfying signal-to-noise ratio, while the repetition time between consecutive pulse sequences was kept long enough (100 ms) to prevent spin-lattice-relaxation effects on the measured spin-echo. The recorded spin-echo was integrated over the measured frequency range, which was swept in 1 MHz steps. The obtained NMR spectra were corrected for the enhancement factor as well as for the ν2-dependence, resulting in relative intensities which are proportional to the number of nuclei with a given NMR resonance frequency.

High-resolution transmission electron microscopy (HRTEM) analysis was performed using an image Cs-corrected Titan 80–300 microscope (FEI) operated at an accelerating voltage of 300 kV. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) imaging and element mapping based on energy-dispersive X-ray spectroscopy (EDXS) were performed at 200 kV with a Talos F200X microscope (FEI) equipped with an X-FEG electron source and a Super-X EDXS detector system. Prior to TEM analysis, the specimen mounted in a double-tilt analytical holder was placed for 10 s into a Model 1020 Plasma Cleaner (Fischione) to remove contamination. Classical TEM specimen preparation based on sawing, grinding, polishing, dimpling, and final Ar ion milling was applied in the case of the sample irradiated with 5 × 1014 ions/cm2.

Results

Magnetometry

The results from SQUID magnetometry performed at 300 K of the irradiated Co2MnSi thin films are displayed in Fig. 1 together with the magnetization loop of a non-irradiated reference sample from the same batch. The diamagnetic background from the substrate and the Ta capping, which leads to a linear decrease (increase) at high (low) field values, has been subtracted. While the coercive field stays constant within error bars (~125–160 Oe), the saturation magnetization decreases noticeably upon increasing the fluence. The biggest step occurs between the non-irradiated reference sample and the sample irradiated with the lowest fluence applied (1013 ions/cm2, 20% reduction). The sample irradiated with the highest fluence (5 × 1015 ions/cm2) exhibits a nearly halved saturation moment. Since magnetic properties of Heusler compounds are strongly linked to the underlying structural order, in the following, we will monitor the structural changes upon irradiation using TEM and NMR.

Figure 1

Magnetization of Co2MnSi thin films at 300 K in fields up to 20 kOe for a non-irradiated reference sample (grey dots and lines) and the irradiated samples with He+ ion fluences ranging from 1013 ions/cm2 up to 5 × 1015 ions/cm2. The data have been normalized to the data of the reference sample in saturation (20 kOe). The diamagnetic background from the substrate and the capping has been subtracted (See text for details).

HRTEM

To study the evolution of the microstructure, HRTEM imaging of the non-irradiated reference and the sample irradiated with a fluence of 5 × 1014 ions/cm2 were performed. In the non-irradiated state, fast Fourier transformation (FFT) analysis (see Supplementary Fig. 4 of supplementary material) points to the presence of the L21 structure, in good agreement to our NMR results (see following section). Regarding the sample irradiated with 5 × 1014 ions/cm2, TEM analysis can differentiate between two sample regions. Evaluating the area close to the MgO substrate, i.e. the lower part of the Co2MnSi film, the diffractograms obtained by FFT are indexed based on the L21 structure in [101] zone axis geometry [Fig. 2(b)], i.e. the same way as for the reference sample. In some parts of the upper Co2MnSi film region, however, particular L21 reflections are strongly decreased in intensity or missing even completely, pointing to an increased disorder. These diffractograms can be better described with the A2 structure, also in [101] zone axis geometry [Fig. 2(c)]. These findings agree well with the NMR results, which found roughly a 50:50 distribution of L21 and A2 structure in this sample. While the NMR measurements cannot determine the location of these structure types within the Co2MnSi layer, HRTEM analysis showed that A2 disorder is mostly present in the upper part of the film. Please note that light grey parts of Fig. 2(a) indicate the presence of oxide columns growing from the surface deep into the film (see supplementary material for details). These inclusions do not interfere with magnetization and NMR measurements, as discussed in detail in the Discussion section.

Figure 2

Cross-sectional HRTEM image (a) and corresponding fast Fourier transforms for the lower part (b) and upper part (c) of the sample irradiated with 5 × 1014 ions/cm2, as obtained from the marked square areas.

NMR

The inset of Fig. 3(a) shows the 59Co NMR spectrum of the non-irradiated reference sample (annealed at 500 °C) in comparison to 59Co NMR spectra of an L21-ordered polycrystalline bulk sample and a Co2MnSi thin film, annealed at 550 °C[33]. Similar to these samples, the spectrum of the non-irradiated reference sample shows a maximum at around 147 MHz, characteristic for the Co2MnSi phase, and a well-defined, reasonably small linewidth, representing a long-ranged, truely L21-ordered 59Co environment (4 Mn and 4 Si atoms as nearest neighbors).

Figure 3

(a) 59Co NMR spectra at 5 K for the non-irradiated reference sample (purple dots). Dotted lines are fits with Gaussian lines, representing the off-stoichiometry. The thick purple line is the sum of these Gaussian lines. The inset shows a comparison of the 59Co NMR spectra of the reference sample (purple dots and line), to a Co2MnSi thin film annealed at 550 °C (black squares and line) and a bulk polycrystalline Co2MnSi sample (red line)[33]. (b) Probability P(n, x) of the binomial distribution function for x = 7.25% (black columns) in comparison to the relative areas of the Gaussian fits to the spectrum (purple columns). (c) Difference between the relative areas of the Gaussian lines and the binomial probability.

It exhibits a low frequency shoulder, stemming from a slight off-stoichiometry of excess Co atoms occupying Mn sites (CoMn antisites)[33]. The amount of CoMn antisites can be estimated by fitting the spectrum with Gaussian lines and comparing their relative areas to a binomial distribution function, resulting from a random atom model for the first shell environment of 59Co[33,38-40]:where P(n, x) is the probability that n sites out of N sites are atoms of concentration x on those N sites. For the first Mn coordination shell (nearest neighboring Mn) of 59Co, N is four and n runs from 0 to 4 in integer numbers. Figure 3 shows such a fit with three equidistant Gaussian lines of equal width (a) and a comparison between their relative areas with a probability function P(n, x) for x = 7.25% (b) for the non-irradiated reference sample. The agreement between the fits and the distribution function is very high, which is further proven in a plot of their difference (c), not exceeding 2%.

Figure 4 shows the evolution of the corresponding 59Co NMR spectra upon the irradiation with He+ ions for all applied fluences. Already for moderate fluences (1013–1014 cm−2), the spectrum broadens slightly, compared to the non-irradiated sample, reflecting an increasing structural disorder. For higher fluences (5 × 1014–1015 cm−2), the broadening gets even more pronounced and the peak frequency shifts to lower values. The spectrum of the sample irradiated with the highest fluence (5 × 1015 cm−2) is completely smeared out over the whole measured frequency range and no distinct peak can be identified. Such a broad NMR resonance line is typical for a completely disordered A2 structure type, where, due to a complete intermixing of all atomic sites, a plethora of different first shell environments exists for the 59Co nuclei, leading to a distribution of many different hyperfine fields[28,41] (see Supplementary Fig. 1 of supplementary material for the L21 and A2 structure types, their corresponding first shell environments and expected NMR spectra).

Figure 4

Normalized 59Co NMR spectra of the irradiated films in comparison to the non-irradiated reference sample.

Figure 5 (A–E) show Gaussian fits to the spectra of the irradiated samples. Three dotted lines indicate the off-stoichiometric L21 phase, similar as already discussed for the non-irradiated reference sample [Fig. 3(a)]. An additional broad Gaussian line (filled light grey line) represents the emerging A2 structure formation. This A2 contribution evolves already for the smallest fluence used. Its percentage increases upon increasing the fluence [see Fig. 5(F)], until reaching 100% for the sample irradiated with the highest fluence (5 × 1015 cm−2, see Fig. 4). The full width at half maximum (FWHM) of the L21 peaks also increases [see Figs 4 and 5(F)]. Due to the increased linewidths, only two Gaussians can be fitted for the L21 order in the last spectrum [Fig. 5(E)]. The spectrum for the highest fluence (5 × 1015 cm−2) is omitted in this plot, since its broad NMR line, ranging over the whole measured frequency range, represents a 100% disordered A2 structure type.

Figure 5

(A–E) 59Co NMR spectra at 5 K for irradiated samples starting with the lowest flux of 1 × 1013 ions/cm2 up to 1 × 1015 ions/cm2. The corresponding fluence is written in the graphs. Lines are fits with Gaussian lines to the spectra. Dotted black lines denote the Gaussians representing the off-stoichiometric L21 ordered regions of the samples, while the filled light grey line corresponds to the A2-ordered regions of the sample. The thick line in the color of the respective data dots is the sum of all Gaussian lines. (F) Irradiation-dependent weight of the A2 structure type (dark green diamonds), FWHM of the L21 phase (orange squares), and the corresponding change in the saturation moment (dark red dots).

Our NMR measurements show that for the considered range of He+ fluences, ion irradiation leads to an increase of structural disorder at the local scale, going along with a strong impact on the macroscopic magnetic properties, even for the lowest fluences applied. We observe a direct crossover from the highly-ordered L21 to the completely disordered A2 structure upon increasing the He+ fluence.

Discussion

Several conclusions can be drawn from the NMR data:

The NMR spectra of all samples show a low-frequency shoulder, which can be fitted assuming three Gaussian lines, whose intensity distributions reflect a slight off-stoi-chiometry of 7.25% excess Co on Mn sites. This kind of off-stoichiometry of an otherwise highly-ordered L21 Heusler structure has been frequently observed in thin films and is typical for stoichiometric sputter targets. It arises from differences in the sputtering rates of individual elements[8,30,33,40,42]. For the investigated samples, the antisite concentration x has been found to be the same throughout the whole irradiation series (see supplementary material).

The L21 lines broaden with increasing ion fluence and an additional, very broad A2 line evolves. These observations imply changes in the local structure of the 59Co atoms on different coordination shells: (i) The probability to have a random distribution of all atoms in higher coordination shells increases. This disorder manifests mainly in the increasing FWHM of the L21 Gaussian lines (dark grey lines in Fig. 5) [(see Fig. 5(F)]. (ii) At the same time, the probability to change also the first coordination shell, which for the non-irradiated sample consists of 4 Mn and 4 Si atoms, increases with increasing fluence. Since within the A2 type (dis)order the total number of different possible atomic configurations in all atomic coordination shells is very large, as is the number of their internal fields, these structural changes lead to the observed additional broad NMR line. For the highest fluences, the majority of Co atoms has a local environment different to those from the non-irradiated sample.

Regarding the type of emerging disorder upon ion irradiation, in principle other structure types such as B2 or DO3 are imaginable. In fact, a similar study on 40 nm thin films of Co2MnSi (originally L21-ordered with some small grains of B2 order) irradiated with He+ ions of much higher energy (150 keV, 1015–1016 ions/cm2) reported an increase of B2 and DO3 disorder upon increasing fluence by means of XRD and HAADF-STEM[43]. However, our NMR spectra do not show any sign of these structure types. For B2, with an intermixing of Mn and Si sites, the NMR spectrum should show additional, well-separated peaks with a fixed spacing on both sides of the central peak[28,41]. In the case of DO3 order (Co-Mn intermixing), peaks with different intensities and spacings on both sides of the central peak would be expected, stemming from the 1st and 2nd coordination shells[2]. The broad additional line in the NMR spectrum of our samples, which evolves upon irradiation, indicates emerging A2 disorder, in excellent agreement with the HRTEM results.

With this study, there are now three different reports on the impact of He+ irradiation on Co2MnSi thin films (which are summarized in Table 1): Gaier et al. reported an improvement of structural and magnetic properties upon the irradiation with 30 keV He+ ions for a certain range of fluences (around 1014 ions/cm2)[23]; a more recent study showed an increase of B2 and DO3 disorder and a reduced magnetization by using 150 keV He+ ions of 1015 to 1016 ions/cm2 [43,44]; while the present work reports a direct crossover from highly-ordered L21 to completely disordered A2 structure (15 keV, 1013–1015 ions/cm2), without inducing other structure types. SRIM calculations (not shown) showed that there is only a slight difference in the average Co displacement upon changing the ion beam energy from 15 to 30 keV He+ ions. The irradiation with 150 keV He+ ions is supposed to lead to bigger differences from local heating due to the larger deposited energy. Apart from the different energies, an explanation for the observed variations might be the different starting conditions. Gaier et al. used B2-ordered films with a certain admixture of A2-disordered regions as starting material. They conclude that the increase of B2 order originates from initially B2-ordered regions, due to Co - Mn and Co - Si exchanges caused by mobile vacancies introduced by the irradiation. Similarly, the L21-ordered films from Abdallah et al. already contained some very small grains of B2 order, which might have favored a (in this case degrading) B2 transformation starting from these initial nucleation sites. The starting material of the present study is highly L21-ordered. The displacement of atoms caused by the incident ion beam produces vacancies, which will stochastically recombine with thermally diffusing atoms resulting in disorder. Since there are no B2 nucleation sites, we expect an arbitrary intermixing of all atoms at all sites, leading to the completely disordered A2 structure, as observed here.

Table 1

Overview of studies on He+-irradiated Co2MnSi thin films, their corresponding experimental conditions and the resulting change of structural properties.

Reference	E (keV)	fluence (ions/cm2)	structure of starting material	structural evolution
this work	15	1013–1015	L21	decrease of structural order (A2)
Gaier et al.[23]	30	1014–1016	B2 + A2	increase of structural order (B2)
Abdallah et al.[43]	150	1015–1016	L21 + B2 grains	decrease of structural order (B2 and D03)

It should also be noted that the reported improvement of B2 structure and magnetic properties in ref.[23] is limited to a certain range of fluences (1–5 × 1014 ions/cm2) for which only a very subtle increase of structural order and saturation moment were observed. For higher fluences also this study found a substantial degradation of structural and magnetic properties.

One of the advantages of the NMR technique is that it allows to probe the local environments of all Co atoms in the film sample in one measurement. (Assuming a relative permeability of 600 (Co) and taking the reported resistivity values of Co2MnSi at 5 K, which range from 4–40 μΩ cm[11,37,45], the skin depth of the penetrating microwave amounts to 400–1000 nm[46], which is far more than the film thickness). While we are probing the full volume of the film without any losses due to film thickness, the NMR method does not reveal from which position in the samples a certain NMR signal is excited, e.g., from Co atoms at the top or at the bottom of the film. Hence, we complemented the NMR results with TEM which can provide spatially resolved information about the local structure at a specific sample position.

According to our HRTEM results, there is a depth-dependent structural order in the irradiated samples: regions next to the surface are mainly A2-ordered, while regions next to the substrate still retain the original L21 structure. In comparison to the SRIM simulations, which showed a rather flat distribution of atomic displacements within the Co2MnSi film (see Supplementary Fig. 2 of supplementary material), this is rather surprising. One may consider this observation to be triggered by the presence of oxide columns growing from the surface deep into the film (down to about 20 nm deep into the film; see Supplementary Fig. 5 of supplementary material and discussion in following paragraph). Here, oxygen may nucleate the vacancies leading to a higher probability for atomic displacements in their vicinity.

Our results underline that structure analysis by NMR, which allows quantitative determination of phase fractions on an integral scale, and by TEM, which enables imaging down to the atomic scale, complement each other very well. A similar combination of NMR and TEM was already successfully used to study the chemical composition of multi-element nanostructures[35].

We now turn to the relation between (local) structure and macroscopic magnetic properties. Note that the absolute value of the saturated moment of our non-irradiated reference sample (2.7 μ/f.u. at 300 K, which translates to 2.8 μ at 5 K[37]) is lower than the theoretically expected and experimentally verified value of 5 μ/f.u.[12,45]. Two issues contribute to the lower absolute value of the saturated moment: (i) Reduced saturated moments have also been reported for other L21-ordered Heusler thin films sputtered from stoichiometric targets, e.g., Co2FeSi (4.5–5 μ instead of 6 μ)[40,47,48]. This difference most likely results from the mentioned off-stoichiometry due to selective sputtering. A polycrystalline bulk sample intentionally prepared with a corresponding off-stoichiometry, namely Co2.1Mn0.9Si, indeed showed a reduced saturation moment of 4.6 μ/f.u. (see Supplementary Fig. 3 of supplementary material). Other theoretical and experimental work on Co2MnSi found comparable effects of Co antisites on the corresponding saturation moments (4.75–4.9 μ, respectively)[14,16,17]. However, the deviation from the expected value for our reference sample is too big (nearly 50%) to be solely explained by a slight Co excess. (ii) In fact, our samples contain considerable amounts of oxide phases (see detailed discussion in supplementary material). In the case of magnetization data, these oxide phases do not contribute to the ferromagnetic signal, but lead to an overestimation of the intrinsic Co2MnSi mass when calculating the magnetization and, thus, to a reduced absolute saturation moment. Importantly, the presence of these oxides does not interfere with the NMR analysis. Measurements of the variations of the local 59Co environment have been made from the same initial state on all samples, namely from the remaining non-oxidized parts of the samples, which gave rise to the ferromagnetic SQUID signals. These are well-characterized Co2MnSi regions, as proven by HRTEM (Fig. 2) and NMR analysis [see inset of Fig. 3(a)]. Both the off-stoichiometry and the overestimation of the intrinsic Co2MnSi mass due to the presence of oxides are expected to be present and on the same order of magnitude in all samples before the irradiation process. Our discussion of the changes in the structure-property relationship of the Co2MnSi thin films upon the irradiation with He+ ions is therefore straightforward and not affected by the presence of oxide inclusions or slight off-stoichiometry.

Finally, we see a clear trend of decreasing moment with increasing ion fluence mimicking the inverse trend of increasing disorder with increasing fluence. These observations are in line with previous studies showing a strong dependence of the magnetic moment on the crystal structure[47,49,50]. Even for moderate fluences (around 1014 ions/cm2), where, based on XMCD measurements, an increasing saturation moment has been reported for Co2MnSi thin films irradiated with He+ ions of similar energy (30 keV)[23], we see a clear trend towards lower saturation moments and a substantial fraction of A2 structure. Hence, ion-irradiation can subtly vary the local environment, thereby revealing structure-property relationships in Heusler compounds.

Conclusion

Combined magnetization, NMR, and TEM measurements of Co2MnSi thin films under the impact of He+ irradiation showed that, even for the lowest fluences applied, the moment in saturation decreases, going along with an increased structural disorder. The disorder manifests itself in a transition from a highly L21-ordered to a completely disordered A2 structure in a sample irradiated with the highest fluence applied (5 × 1015 He+ ions/cm2). The direct transition from L21 to A2 has been revealed by NMR measurements, which could deduce the amounts of L21- and A2-ordered phases in each of the thin films. The NMR measurements also revealed a slight off-stoichiometry of 7.25% Co excess, which is constant throughout the films, indicating that it originates from their preparation process. Complementary HRTEM analyses added local information of the distribution of L21 and A2 phases in the sample, revealing that the A2 disorder is mainly located at the upper part of the films in vicinity to the film surface.

SUPPLEMENTARY MATERIAL - Structure-property relationship of Co$_2$MnSi thin films in response to He$^+$-irradiation